joshuasiagian/indosum

INDOSUM is a benchmark dataset for Indonesian text summarization containing approximately 19,000 detokenized news articles from Shortir.com (a news aggregator site). The dataset consists of simplified article-summary pairs.

Dataset Details

Languages

• Indonesian (ind)

Supported Tasks

• Text Summarization (abstractive and extractive)

Dataset Usage

Using datasets library

from datasets import load_dataset

# Load the default configuration (all folds combined)
dataset = load_dataset("indosum")

# Load a specific fold for cross-validation
dataset = load_dataset("indosum", "fold0")

# Access the splits
train_data = dataset["train"]
val_data = dataset["validation"]
test_data = dataset["test"]

Available Configurations

• default: Same as fold0 - uses the first fold (~19K articles)
• fold0: Fold 0 data (train.00, dev.00, test.00) - ~14,262 train / ~750 val / ~3,762 test
• fold1: Fold 1 data (train.01, dev.01, test.01) - ~14,263 train / ~749 val / ~3,762 test
• fold2: Fold 2 data (train.02, dev.02, test.02) - ~14,290 train / ~747 val / ~3,737 test
• fold3: Fold 3 data (train.03, dev.03, test.03) - ~14,272 train / ~750 val / ~3,752 test
• fold4: Fold 4 data (train.04, dev.04, test.04) - ~14,266 train / ~747 val / ~3,761 test

Note: The dataset provides 5-fold cross-validation splits of approximately 19,000 unique articles. Each fold contains the same articles organized differently for cross-validation. For most use cases, use the default configuration. For cross-validation experiments, iterate through fold0 to fold4.

Data Format

Each example in the JSONL files contains:

• paragraph: Full article text detokenized from the original tokenized paragraphs
• summary: Gold summary detokenized into a single string

The original metadata fields and tokenized sentence structure are removed to keep the dataset focused on summarization.

Dataset Homepage

https://github.com/kata-ai/indosum

Dataset Version

Source: 1.0.0. SEACrowd: 2024.06.20.

Dataset License

Apache License, Version 2.0

Citation

If you are using the Indosum dataloader in your work, please cite the following:

@INPROCEEDINGS{8629109,
  author={Kurniawan, Kemal and Louvan, Samuel},
  booktitle={2018 International Conference on Asian Language Processing (IALP)}, 
  title={Indosum: A New Benchmark Dataset for Indonesian Text Summarization}, 
  year={2018},
  volume={},
  number={},
  pages={215-220},
  doi={10.1109/IALP.2018.8629109}}


@article{lovenia2024seacrowd,
    title={SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages}, 
    author={Holy Lovenia and Rahmad Mahendra and Salsabil Maulana Akbar and Lester James V. Miranda and Jennifer Santoso and Elyanah Aco and Akhdan Fadhilah and Jonibek Mansurov and Joseph Marvin Imperial and Onno P. Kampman and Joel Ruben Antony Moniz and Muhammad Ravi Shulthan Habibi and Frederikus Hudi and Railey Montalan and Ryan Ignatius and Joanito Agili Lopo and William Nixon and Börje F. Karlsson and James Jaya and Ryandito Diandaru and Yuze Gao and Patrick Amadeus and Bin Wang and Jan Christian Blaise Cruz and Chenxi Whitehouse and Ivan Halim Parmonangan and Maria Khelli and Wenyu Zhang and Lucky Susanto and Reynard Adha Ryanda and Sonny Lazuardi Hermawan and Dan John Velasco and Muhammad Dehan Al Kautsar and Willy Fitra Hendria and Yasmin Moslem and Noah Flynn and Muhammad Farid Adilazuarda and Haochen Li and Johanes Lee and R. Damanhuri and Shuo Sun and Muhammad Reza Qorib and Amirbek Djanibekov and Wei Qi Leong and Quyet V. Do and Niklas Muennighoff and Tanrada Pansuwan and Ilham Firdausi Putra and Yan Xu and Ngee Chia Tai and Ayu Purwarianti and Sebastian Ruder and William Tjhi and Peerat Limkonchotiwat and Alham Fikri Aji and Sedrick Keh and Genta Indra Winata and Ruochen Zhang and Fajri Koto and Zheng-Xin Yong and Samuel Cahyawijaya},
    year={2024},
    eprint={2406.10118},
    journal={arXiv preprint arXiv: 2406.10118}
}